LeNet5-MNIST-PyTorch

This is the simplest implementation of the paper "Gradient-based learning applied to document recognition" in PyTorch.

Have a try with artificial intelligence!

Feel free to ask anything!

Activation Difference!

Note that this repo's implementation uses MaxPooling and ReLU instead of AvgPooling and Sigmoid activation.

Requirments

Python3
PyTorch >= 0.4.0
torchvision >= 0.1.8

Usage

$git clone https://github.com/ChawDoe/LeNet-5-MNIST-PyTorch.git  
$cd LeNet5-MNIST-PyTorch  
$python3 train.py  

model will now run on GPU if available

Hint

This repo includes the mnist dataset.

Accuracy

Average precision on test set: 99%